Computational Electromagnetics


Current efforts are directed towards the development of high performance portable electromagnetics codes for scalable architectures as part of the Department of Defense High Performance Computing Modernization Program's (DoD HPCMP) Common High Performance Software Support Initiative (CHSSI). The effort is part of the Computational Electromagnetics and Acoustics (CEA) Computational Technology Area (CTA) lead by Dr Joseph J.S. Shang at Wright Laboratory at the Wright-Patterson Air Force Base.

For the scalable code development the first phase consists in the development of an optimized uniprocessor Fortran-90 code as a basis for a high performance message passing code for distributed memory scalable architectures, such as the IBM SP-2 and the Cray T3E. An MPI version will be produced for portability and an Active Message version will be produced for evaluation of the benefits of a leight weight interprocessor communication protocol for production quality computational electromagnetics codes. In addition, a HPF (High Performance Fortran) version will be produced for the evaluation of HPF compilers as they appear and the quality of their optimization features improve to offer competitive performance for programming in a shared address space.

The scalable code currently being developed is a multi-block finite-volume code based on high resolution algorithms devised and analyzed by J.S. Shang et. al. The high resolution algorithms allow for significantly fewer grid points per wavelength than previous algorithms. The developed software will extend the application envelope in frequency spectra accordingly and achieve a numerical accuracy, when needed, up to a dynamic range of 60 db. The code use generalized curvilinear coordinates. The code will accommodate complex configurations such as aircrafts, ship hulls, armored vehicles, buried mines, and antenna arrays.

We also expect to develop multipole-like scalable computational electromagnetics codes as part of this effort.


The overall effort can be divided into several projects or tasks, each suitable for thesis work.